CARBON-MOTIVATED BORDER TAX ADJUSTMENT: POLICY DESIGNS COMPARISON

Transcription

1 CARBON-MOTIVATED BORDER TAX ADJUSTMENT: POLICY DESIGNS COMPARISON Paola Rocchi *, Iñaki Arto**, Jordi Roca*, and Mònica Serrano* * Faculty of Economics and Business, University of Barcelona, Spain ** bc 3 Basque Centre for Climate Change PLEASE DO NOT QUOTE (Preliminary version April 2015) Abstract: The following analysis focuses on carbon-motivated border tax adjustments (), which are tariffs applied by countries implementing carbon control policies imposed on products imported from abroad. In particular we focus on s policy design, since s can be computed considering emissions embodied in imports, or referring to emissions avoided through imports. Using the WIOD database, we simulate through a multi-region and multi-sector analysis what tariffs system should be applied to products imported in Europe to compensate an European CO 2 emissions taxation, considering embedded emissions or avoided emissions. To know for which countries and sectors the method used is critical can help to understand and to add information to the political debate on it. Furthermore, an important novelty of our analysis is that we estimate avoided emissions not only using the traditional domestic technology assumption, but also using a more appropriate approach that considers the physical quantities of imported goods. 1

2 1. Introduction Within the framework of emission control instruments, there is currently an important debate regarding carbon-motivated border tax adjustment () and its possible role to complement emission reduction policies. s are a trade instrument, consisting in tariffs applied by countries that are pricing carbon emissions, and they are aimed at preventing the main drawbacks of unilateral emission control mechanisms. If implemented, these tariffs would be imposed on products imported from all the countries that are not applying the carbon control policy in order to level the carbon playing field, 1 compensating for the loss of competitiveness that a carbon tax may imply for domestic producers, and avoiding possible emission leakage involved in unilateral emission reduction policies (Lockwood and Whalley 2010, Horn Sapir 2013). 2 As Mattoo et al. (2009) and Kuik (2010) point out, the viability of this tool has already been considered at political level and included in policy proposals of countries like United States (US) and Europe. Indeed, in the 2009 US proposal for implementing an emission trade mechanism - the American Clean Energy and Security Act 3 - the border adjustment measure was included as a competitiveness measure to ensure the equal distribution of costs in the absence of an international agreement limiting emissions. Looking at the European Union (EU), the revised Emission Trading Scheme (ETS) directive (European Parliament and Council, 2009) asks for an assessment of the impact of carbon leakage on Member States and for the inclusion of importers in the Community scheme. Also international trade authorities such as the World Trade Organization (WTO) have already reckoned the relevance of s (see WTO 2009 and Hillman 2013). One of the critical issues related to the implementation of s is its policy design, since tariffs can be compute through different methods. In particular, there are two main approaches to calculate a. One way is to compute the tariffs based on the emissions embodied in each imported products. Alternatively, the tariff could be computed on the emissions avoided through imports: in this case the tariff applied to any imported product would be based on the emissions embodied in the same good produced domestically. The debate revolves around three implications of the different policy designs. The first one is its legal acceptability as a trade instrument, considering the international legal framework established by the WTO. The second dimension is the environmental effectiveness, and the third one is the political feasibility in terms of practical implementation. On the one hand, tariffs calculated on avoided emissions are more justifiable as a trade policy considering that the WTO regulation detailed by GATT (1994) would accept indirect taxes related to products and not direct taxes related to producers or to their income (Mattoo et al. 2009, Hillman 2013). Moreover, s on avoided emissions would also be easier to be implemented, because they would imply no discrimination among exporting countries. On the other hand, they could be considered less effective as an environmental policy: indeed, tariffs on avoided emissions would not give any incentive for exporting countries to implement more environmentally- 1 This expression is often used in s literature. See, for example, Houser et al. (2008). 2 To better justify BTAs, Horn and Sapir (2013) refer to international externalities that arise when countries combat emissions unilaterally. Indeed, countries implementing unilateral climate policy face the full costs of their abatement efforts, receiving only part of the benefits that are spread across the world. As a result, they will typically choose sub-optimal climate policies exposing each other to more climate damage than would be internationally efficient, that is, exposing each other to international externalities. 3 The American Clean Energy and Security Act (American House of Representative, 2009) was a US energy bill that, if approved by the Senate too, would have established an emission trading mechanism similar to the European Union Emission Trading Scheme. 2

3 friendly technologies. Then, if the main characteristic of this policy is its environmental purpose, s calculated on embodied emissions would have stronger justification. Focusing on the different methods of designing a system, in the following analysis we simulate what tariffs system should be applied to the products imported in Europe in order to compensate an equivalent taxation on CO 2 emissions implemented within the EU. We focus in particular on Europe due to its position about carbon pricing. Indeed, on the one hand, the EU debate on pricing carbon emissions has a long history started in the early 1990s, and the EU is already implementing different policies for emission control such as the ETS or the European Energy Tax Directive (ETD), a tax on the use of energy products aimed at reducing emissions. On the other hand, these main policies implemented so far are still weak or poorly performing, 4 and there are ongoing political debates to strengthen them in order to reach the challenging environmental targets Europe has set itself. 5 Despite the political difficulties in advancing in carbon taxation in the EU, we believe it is important to revive the debate on implementing a harmonised EU carbon tax as a powerful climate change tool to reduce emissions. For this reason, since s have already been feared as possible for complementing a carbon tax, it seems important to analyse all the critical issues that they would imply, among them what method should be used to compute them. More in detail, the aim of our work is to analyze what differences would exist between s calculated on embodied emissions and CBAs calculated on avoided emissions, by using a multi-region and multi-sector analysis that permits to show to what extent the two policy designs would affect differently the countries or sectors that have trade relationships with Europe. To know for which countries and sectors the method used is critical can help to understand and to add information to the political debate on it. There is an already vast literature on BTAs (see Ghosh et al for a survey); anyway, due to the scope of our analysis, we focus in particular on the papers that compare different methods of computing the tariffs system (Mathiesen and Maestad 2002, Kuik and Hofkes 2009, Mattoo et al. 2009, Burniaux et al. 2010, Lin and Li 2011, Böhringer et al. 2012, and Elliott et al. 2012), or the papers that use a multi-region multi-sector analysis to determine s rates (Mattoo et al. 2009, Atkinson et al. 2011, Dissou and Eyland 2011, Böhringer et al. 2012, Elliott 2012, Ghosh et al. 2012, and Schenker et al. 2012). Among them, only three papers - Mattoo et al. (2009), Böhringer et al. (2012) and Elliott et al. (2012) - consider both issues together, using an input-output (IO) approach to compute and comparing different s designs. Indeed, except from these three studies, all the analyses that compare different s designs do not use an IO approach to compute emissions embodied and emissions avoided. Instead, they consider the direct emissions of the different sectors through sectors emission intensity, using in particular the emission intensities of target countries for computing embodied emissions, and the emission intensities in the countries implementing the policy for computing avoided emissions. Alternatively, Burniaux et al. (2010) consider for each sector the direct emissions (computed through emission intensities) and the indirect emissions that are the sum of direct emissions and emissions embodied in sectors electricity use. Differently from this literature, we believe that considering all the emissions embodied in different goods through an IO approach is more appropriate. In fact, any carbon taxation could apply, domestically, to all sectors; in this case the final additional cost for any 4 Indeed, the European ETD currently in force fixes very low tax rates for the most part of fuel uses and does not explicitly tax energy products according to their carbon emissions; looking at the ETS, during the last years the carbon price has been too low to give a strong price signal. 5 In addition to the ETS Directive quoted above (European Parliament and Council, 2009), in 2001 the European Commission already proposed to modify the ETD in order to introduce an explicit carbon tax component (European Commission, 2011). 3

4 domestic good would depend on the direct emissions produced by its sector, but also on the pollution emitted when making all inputs needed to produce it. So, since the aim of s is to apply the same treatment to domestic and imported products, considering exporting countries that are not imposing the carbon taxation, many energy-intensive sectors such as electricity but also transportation are producing mostly non-tradable goods, but their emission content should be taken into account if they are used as an input to produce tradable goods that are then imported by the country applying the carbon tax. Looking at the papers that propose a multi-region multi-sector analysis to determine s rates, they usually do not offer a comparison between different policy designs, because they focus on different issues related to s. In particular, Atkinson et al. (2011) compare the total emissions embodied in domestic production with the emissions embodied in consumption. They use IO information to find out the average tariffs but just considering the emissions embodied in imports through a bilateral trade IO approach. 6 They find that the main exporters of virtual carbon are China and Russia, while the main importers are EU, USA, and Japan; tariff rates would widely vary by country and by sector, the large developing countries showing tradable sectors such as chemicals or mineral products that would be particularly hit by s. Dissou and Eyland (2011) analyse different BTA recycling methods; they find that using s proceeds to support energy-intensive industries improves their competitiveness. Ghosh et al. (2012) focus on efficiency and distributional consequences of s calculated only on CO 2 emissions or s calculated on different greenhouse gases (GHG); the main finding is that s distributional implications are smaller under a broad-based GHG policy. Schenker et al. (2012) analyse s in terms of output variation, welfare effect, carbon leakage, and trade composition. They compare a full IO approach to an IO approach corrected for imports from regulating countries (Europe, Australia, Japan, Korea, and Taiwan), and they find that the two methods do not imply strongly different results. Analyzing more in detail the three papers that consider both issues together, Mattoo et al. (2009) assess the different impact of s computed on embodied emissions and s computed on avoided emissions on several target countries, assuming unilateral emissions reductions of 17% by 2020 in high income countries (EU, USA, Japan, and other United Nations Framework Convention on Climate Change (UNFCCC) Annex-I countries). They use a computable general equilibrium model based on 2004 GTAP data. The main finding is that, on the one hand, s on embodied emissions would imply average tariffs for India and China of over 20% and would depress manufacturing exports between 16 and 21%. On the other hand, s on avoided emissions would address the competitiveness problems without so many damages for exporting countries. Böhringer et al. (2012) compute the efficiency impact of different s designs, analyzing three different regulating coalitions: Europe, UNFCCC Annex-I regions except for Russia, or a broad coalition that includes China. They simulate a unilateral cap at 80% of the abating regions emissions. s vary among three dimensions: embodied carbon coverage (direct, direct and electricity-related, or total emissions), sector coverage (energy-intensive trade-exposed goods, or all goods), tariff rate differentiation (country- and sector-specific, or sector-specific tariffs). They find that systems more likely to comply with international law yield very little in terms of carbon leakage and efficiency. Elliott et al. (2012) analyze the extent of emission reduction of a wide range of carbon tax schemes in Kyoto protocol Annex-B countries, the expected carbon leakage, and the effect of s, simulating both s on embodied emissions as well as s on emissions related to production technologies in importing countries. Using 2004 GTAP data through a computational general equilibrium model, they show the 6 Through this approach they consider only the carbon added by the exporting countries, and not the emissions embodied to produce the inputs that the exporting countries are importing from abroad. 4

5 importance of global taxes: carbon taxes only in Kyoto protocol Annex-B have low potential to reduce emissions. They also find that s on avoided emissions can be significantly inferior at reducing leakage than s on embodied emissions, mainly due to the lack of incentives for foreign producers to adapt less-polluting technologies. Our work follows this research line, with some differences. First, instead of focusing on the broad effects of s in terms of output, competitiveness or environmental goals as the previous literature does, we propose static analysis to show what tax level each policy design would imply. Comparing our proposal to Mattoo et al. (2009), Böhringer et al. (2012) and Eliott et al. (2012), these works do not provide an analysis of the different s designs at a sector level, because they address competitiveness, environmental, or efficiency concerns in global terms using computational general equilibrium models. On the contrary, we propose a sectorbased country-based analysis. We think that this is an important contribution because the intensity but also the spread or concentration of the effect of BTAs designs in different countries and different sectors is an additional element to assess the feasibility of this policy. Moreover, while the previous studies use the GTAP database, we employ WIOD data that are better suited with the scope of our analysis. Furthermore, our analysis explores an additional methodological issue about how to calculate the avoided emissions. Indeed, the estimation of avoided emissions is equivalent to compute the emissions of imported products applying the so called domestic technology assumption (DTA). As we have analysed in a previous article, the usual way of estimating emissions according the DTA has a problem that could significantly bias the outcomes (Arto et al., 2014), because the implicit assumption is that prices of imported goods are equal to prices of the same products produced at home. For this reason in this paper we estimate avoided emissions correcting for the differences in prices of imported and domestically produced goods using trade data in physical units as it is explained in Arto et al. (2014). Finally, we summarize the results of our analysis through two different indices. The first one offers a synthetic measure of the s effect for each target country, considering the quantity exported by that country to the EU, in order to compare the s effect at a country level. The second index measures the effect of s within each European country considering its trade relations with non-european countries. Since one drawback of s is their negative effect for domestic producers using foreign inputs and for domestic consumers buying foreign products, this exercise permits to show what European countries would be most affected by s due to their international trade structure, so it permits to show some preliminary information also regarding the different obstacles this policy could meet within Europe. 2. Method The input-output framework offers adequate instruments to the scope of our analysis. Indeed, the aim of a border tax system is to apply a fiscal treatment to imported products similar to the one applied to domestic products. When a carbon tax is implemented, the fiscal load charged on a domestic good is related not only to the pollution that a sector emits to make it, but also to the emissions contained in the inputs used to produce it, because also the sectors producing the inputs would be subject to the same tax. So, if a border tax wants to apply the same treatment to foreign products, it has to account for the input structure and the input emissions content of foreign products too, and the input-output analysis is a framework precisely constructed to do it. 5

6 As Arto et al. (2014) point out, the literature has extensively shown that two different methods are needed to compute emissions embodied in trade and emissions avoided through trade. Emissions embodied in trade can be traced out using an environmentally-extended multiregional input-output (ee-mrio) model. In a world consisting in c countries, each of them composed of n sectors, the ee-mrio model considers the equivalence between the total output of any sector i in country r and how that output is consumed, domestically or abroad, by any economic sector j as an intermediate input, or by final users as final output : n c c r rs rs i ij i j 1 s 1 s 1. In matrix x x f terms the same expression becomes, being the n c vector of sectoral output, the n c n c matrix of all the inter-country inter-sector deliveries, and the n c vector of final demand. The equivalence can be then transformed in. is the matrix of input coefficients representing the world input structure, where each element is obtained as. Adding information on emissions per sectoral unit obtained dividing the total sectoral emissions over the output produced by each sector, it is possible to compute the emissions embodied in each unit exported. Anyway, since s would be tariffs applied to foreign products, instead of computing the embodied emissions per sectoral unit, we compute the embodied emissions per each one of the products any country is producing. The m c -dimensional vector of emission content of the total export for each product from any country can be indeed calculated as : is a one s vector of appropriate dimensions, is the matrix containing in the diagonal the n c emissions per sectoral unit, while is a n c m c matrix containing the total export desegregated by sector and by product. The embodied emissions per product unit are equal to the vector of total emissions of exports by product divided by the vector of total exports by product ( ). The method needed to compute avoided emissions is an environmentally-extended singleregion input-output (ee-io) model, applying the so-called domestic technology assumption (DTA): 7 through this model we calculate the amount of emissions that would have been contained in a domestic product if all its inputs were produced domestically, with the technology available domestically. Applying the DTA is indeed useful in this context, as shown by the following simple example. Considering a cloth imported from India, the border tax on avoided emissions applied to it would be equal to the carbon tax charged on a European cloth. If the Indian cloth is then used to produce a blouse internally, the tax burden on the blouse produced in Europe using Indian cloth has to be equal to the tax charged on a European blouse made with European cloth in order to treat domestic and foreign products in the same way. For the same reason, in a regime of duties based on avoided emissions, the same fiscal load would be then applied also to a shirt directly imported from India. So, considering only one single region with sectors, and applying the DTA, the n-dimensional vector of emissions avoided by importing any product from abroad would be. is a one s vector of appropriate dimensions, is the matrix containing in the diagonal the emissions per sectoral unit. represents the matrix of total input coefficients, sum of the domestic input coefficient matrix and the matrix of imported input coefficient :. 7 As explained in Arto et al. (2014), we maintain the acronym DTA often used in literature, although, previously, it was mostly used in a different context, to describe the foreign technology when MRIO data were not available. 6

7 is a matrix containing the total imports desegregated by sector and by product. The avoided emissions per product unit are equal to the vector of total emissions of imports by product divided by the vector of total imports by product ( ). Since input-output data are expressed in monetary terms, when the DTA is used to compute avoided emissions an important issue related to the units considered arises, as it is extensively shown in Arto et al. (2014). When we calculate the emissions applying the DTA in monetary terms, we actually estimate the emissions produced by making the imported goods with the domestic technology, but we consider the value of imported goods instead of the physical quantities imported. Due to differences in prices among countries, it can be the case that the monetary value of imports would correspond to a different monetary value if the products imported would have been produced domestically. Or, in other terms, to calculate the emissions that would have been produced to make domestically all the products usually imported is important to adjust for the level of prices, or to deflate the value of imports in order to account for international price differences. So, following the suggestion of Arto et al. (2014), we compute emissions avoided by trade applying both, the monetary DTA and the physical DTA adjusting for price differences across countries. Appendix 1 explains in detail how deflators are computed. Finally, two indices are computed, to show the effects of any s design on different countries, both European or target countries. In particular, for every European country we compute the cost of any s design as a percentage of the total value imported by that country from all the non- EU countries. In a similar way, for every non-eu country we compute the total load any s system would imply, as a percentage of the total value exported from that non-eu country to all the European countries. 3. Data description The analysis requires three main data, which are MRIO tables, CO 2 emissions data and international trade data in physical and monetary units. For MRIO and environmental data, we use the database made available by the WIOD project since April 2012 and updated in November 2013 (WIOD, 2013). The WIOD database consists of four main time series: world input-output tables and international supply and use tables (WIOT-ISUT); national input-output tables and national supply and use tables (NIOT-NSUT); socio-economic accounts (SEA); environmental accounts (EA). In particular, for MRIO tables, the study considers the information contained in the first series, WIOT-ISUT. These data, available for the years , refer to 27 European counties, 13 other major countries in the world and all the remaining regions aggregated in a single rest of the world region. 8 Among the different tables contained in the WIOT- ISUT series, 9 the world input output table at current prices for the year 2009 is used. 10 It is a 8 See Appendix 2 for the complete list of countries. 9 The full set of the WIOT-ISUT tables contains international supply and use tables at current and previous year prices, with use split into domestic and import by country (35 industries by 59 products), world input output tables at current and previous year prices (35 industries by 35 industries), and interregional input output tables for 6 regions (35 industries by 35 industries). The used classification is the National Classification of Economic Activities (NACE) Rev 1.1 (Eurostat, 2002), for industries, and Classification of Product by Activities (CPA) (Eurostat, 2008), for products. 10 The year is chosen due to the time availability for data on CO 2 emissions. Data are expressed in million of US dollars. To convert them in euro, annual data about exchange rates made available by Eurostat (2014) are used. 7

8 symmetrical table industry by industry, offering a desegregation of about 35 sectors for each country. 11 This industry-type table is estimated under the assumption of fixed product sales structure, which states that each product has its own specific sales structure, irrespective of the industry where it is produced. Finally, to get information desegregated by product, we also use the 2009 international supply and use tables at current prices, where information is available for 59 CPA products. 12 In WIOT_ISUT data are expressed in monetary terms (millions of dollars). For CO 2 emissions data, we use the EA made available by WIOD. WIOD satellite accounts record data about use of energy, emission of main greenhouse gases, emission of other main air pollutants, use of mineral and fossil resources, land use, and water use. This satellite accounts have the same sector breakdown (35 sectors) and geographical coverage (40 countries including 27 European countries, and the remaining regions aggregated in a single region Rest of the World ) as the WIOT- ISUT series, and they are available for the years Air emission accounts include CO 2 emissions (in 1000 tonnes) desegregated by sector and energy commodity, and non-co 2 emissions (in tonnes) by sector. CO 2 emissions include both energy-related air emissions that result directly from the use of energy through fuel combustion and non-energy related air emissions that are not directly related to the combustion process, such as industrial process including mineral, chemical and other production sectors, agriculture including manure, agriculture soils and field burning and waste. 13 Finally, we use data on international trade from the database COMEXT made available by Eurostat (Eurostat, 2015), using in particular data recorded following the 2002 classification CPA. This database contains statistics collected by member states on merchandise trade among EU countries and between member states and global partners for the period Data are available for 283 trading partners and 881 products categories, and they are expressed in monetary terms (euro) as well as in physical term (kilograms). In particular, we use the information for the 14 non-european countries available in WIOD too (see Appendix 2 for a list), and information on 217 product Results The simulation we propose shows the s rates that would apply on products imported from non-eu countries if Europe adopts a system to complement domestic carbon taxation equal to 20 euro per tone of CO 2 emitted. We fix the domestic carbon tax at this level because it is the CO 2 tax rate lately proposed by the European Commission - but not approved by the European Parliament to implement a specific carbon component in the taxation on the use of energy products (European Commission, 2011), so it represents a tax level that could be realistically implemented by the EU. The rates are calculated by product, and they are computed based on the CPA disaggregation available in WIOD data. So, the rates shown in this analysis are indeed average tariffs, assuming a unique homogeneous good for each CPA classification. 15 Although data are disaggregated in 59 different categories, we focus our analysis on the 22 manufactured products, excluding agricultural products, raw material and services. We indeed assume that services would not be included in a system. Moreover, we also assume that tariffs would not be imposed on 11 See Appendix 3 for the complete list of sectors. 12 See Appendix 4 for the complete list of products available in WIOD. 13 The matrix of CO 2 emissions is obtained by applying CO 2 emissions coefficients to emission relevant energy use data and then adding process-based emissions. 14 See Appendix 5 for a complete list of the 217 products used from COMEXT (Eurostat, 2015). 15 Actually each CPA category aggregates a wide variety of products. Reason why, starting from the results found in this analysis, a possible extension of this work could be a more desegregated analysis focused on the CPA products that would be charged most under a scheme. Anyway, this is not possible with WIOD data that instead permit a comparison among several countries and a multi-regional analysis. 8

9 products that do not have an equivalent good produced domestically, because the aim of a system would be to apply to foreign products the same fiscal regime implemented to domestic goods. Since for agricultural products and raw material the disaggregation available does not permit to distinguish between products that Europe is importing but also producing domestically from that products or raw material that the EU does not own and needs to import from abroad, we exclude them from the analysis. Table 1 shows rates computed on embodied emissions and rates computed on avoided emissions for each non-eu country. The rates found are directly related to emissions produced when making any product. Tax rates on embodied emissions vary by country because each country has a different technology for each product; regarding tax rates computed considering avoided emissions, while the emissions avoided by importing a product are the same independently of what country the product is imported from, the rates applied on the value of import vary among countries due to international differences in prices. We compare first the two different tax designs considering all the products and all the countries in aggregate terms, finding that, as a global effect, tariffs would be higher in a system based on embodied emissions. To analyze the results in aggregate terms, in figure 1 we distinguish among products that would be more strongly affected through tariffs higher than 2%, products that would be mildly affected (with tariffs between 1% and 2%) and less affected products (with tariffs lower than 1%). In a system based on embodied emissions (Figure 1.a) for 41% of the totality of the 308 products considered (22 products for each of the 14 countries) the tariff applied would be more than 2%, while for the 42% of the products would be between 1% and 2%. In a s system calculated on avoided emissions (Figure 1.b) these percentages would be, respectively, 5% and 18%. Figure 1. Percentage of products based on their tariff size, under the two different s designs 41% 17% 42% 5% 18% 77% 32% 16% 52% tariff < 1% 1%< tariff <2% tariff >2% a. Embodied emissions b. Avoided emissions (monetary) Source: own elaboration. c. Avoided emissions (adjusted for price differences) This strong difference between s calculated on embodied emissions and calculated on avoided emissions is partially explained through international differences in prices. Adjusting for price differences (Figure 1.c) reveals that the most polluting products or the products produced by the most polluting countries - are, on average, cheaper than European products, that implies that, after deflating data, the percentage of products strongly affected would be higher compared with the percentage found without adjusting for price differences (16% instead of 5%). Also mildly affected products, as the strongly affected ones, would be proportionally more when adjusting for price differences (32% instead of 18%). So, if on the one hand the impact of a system based on embodied emissions would be stronger, on the other hand this first result reveals that is necessary to take into account international price differences to avoid biased outcomes. 9

10 Table 1. s by product for any non-eu country, in percentage, corresponding to a 20 euro/ CO 2 tonne European carbon tax NEW DEFLATORS AE * EE ** AUS BRA CAN CHN IDN IND JPN *** AE d EE AE d EE AE d EE AE d EE AE d EE AE d EE 15 Food products and beverages Tobacco products Textiles Wearing apparel Leather and leather products Wood and products of wood and cork Pulp, paper and paper products Printed matter and recorded media Coke, refined petroleum products Chemicals, chemical products Rubber and plastic products Other non-metallic mineral products Basic metals Fabricated metal products Machinery and equipment n.e.c Office machinery and computers Electrical machinery Radio, television and comm. eq Medical and optical instruments Motor vehicles, trailers Other transport equipment Furniture; other manufactured goods Average tariff by country Source: own elaboration. Non-EU countries: AUS: Australia; BRA: Brazil; CAN: Canada; CHN: China; IDN: Indonesia; IND: India; JPN: Japan. * Carbon border tax calculated on the emissions avoided by Europe through trade. ** Carbon border tax calculated on emissions embodied in the products imported in Europe. ***Carbon border tax calculated on the emissions avoided by Europe through trade, adjusting for international prices differences. - Product not exported to Europe. AE d 10

11 Table 1. (Continuation) s by product for any non-eu country, in percentage, corresponding to a 20 euro/ CO 2 tonne European carbon tax KOR MEX RUS TUR TWN USA ROW AE * EE ** *** AE d EE AE d EE AE d EE AE d EE AE d EE AE d EE AE d 15 Food products and beverages Tobacco products Textiles Wearing apparel Leather and leather products Wood and products of wood and cork Pulp, paper and paper products Printed matter and recorded media Coke, refined petroleum products Chemicals, chemical products Rubber and plastic products Other non-metallic mineral products Basic metals Fabricated metal products Machinery and equipment n.e.c Office machinery and computers Electrical machinery Radio, television and comm. eq Medical and optical instruments Motor vehicles, trailers Other transport equipment Furniture; other manufactured goods Average tariff by country Source: own elaboration. Non-EU countries: KOR: Korea; MEX: Mexico; RUS: Russia; TUR: Turkey; TWN: Taiwan; USA: United States; ROW: Rest of the World. * Carbon border tax calculated on the emissions avoided by Europe through trade. ** Carbon border tax calculated on emissions embodied in the products imported in Europe. ***Carbon border tax calculated on the emissions avoided by Europe through trade, adjusting for international prices differences. - Product not exported to Europe. 11

12 4.1. Analysis at the product level Analyzing the results at a product level, the difference between the two approaches found in aggregate terms still exists, although some anomalies can also be found. In this section we propose a product-based analysis, looking first at the results obtained assuming s on embodied emissions, and then comparing them with s computed on avoided emissions. If we look at the tariffs obtained simulating a system based on embodied emissions, the CPA category mostly affected would be other non-metallic mineral products (26); 16 for these products, the average rate would be higher than 2% in all the 14 non-eu countries considered, being particularly high (more than 10%) for products imported by China, Indonesia, India, Russia, and Taiwan (respectively, 10.1%, 12.3%, 12.9%, 12.8%, 12.3%). Among all the manufactured products analyzed, these products are the ones whose emissions depend most on domestic technologies: for all the countries considered except for Canada the emissions produced by each country are at least the 90% of total emissions embodied in these products. But, while for Indonesia and India these emissions are largely due to the technology of the sector producing other non-metallic mineral products, in the case of China and Russia an important share of emissions (32.1% and 32.8%) are embodied in the electricity need to produce them. As regards Taiwan, one fifth of embodied emissions are related instead to the extraction of raw material. Other products that would be particularly affected are basic metals (27), and fabricated metal products (28). For these products, rates would be high in particular for Russia (10.2%), India (8.2%, 7.4%) followed by Indonesia (6.7%), and China (6.3% and 6.2%). China, India, Russia and Taiwan would also have the highest rates on other energy-intensive products coke, refined petroleum products (23), chemicals, chemical products (24). For all these products, the analysis reveals a pattern of embodied emissions very similar to the one described for other non-metallic mineral products : on average, for the countries analyzed, roughly the 80% of embodied emission are produced domestically. For basic metals (27), and fabricated metal products (28) produced in Russia and Indonesia emissions are mainly due to the intensive use of energy of the producing sectors; for Chinese and Indian products belonging to these classifications, and for coke, refined petroleum products (23) or chemical products (24) from the countries listed before, roughly half of the emissions embodied are due to the electricity needed to produce them. Some of these products would also have rates higher than 2% when imported from Australia, Canada, Korea, Mexico, Turkey and USA, but in this case the contribution to emissions of the electricity sector would be much lower. There are finally other products that would be taxed most, but just when produced and imported by few countries, in particular China, India, and Russia. For these three countries, many products would be taxed with rates higher than 3%: wood and products of wood and cork (20), pulp, paper and paper products (21), printed matter and recorded media (22), machinery and equipment (29), office machinery and computers (30), electrical machinery (31), radio, television, and communication equipment (32), medical and optical instruments (33), motor vehicles, trailers (34), other transport equipment (35). Once again, what these three countries have in common is that they rely most on domestic production (for all these products more than 90% of embodied emissions are produced domestically) and they also have in common a relevant role of the electricity sector in creating the emissions embodied in these products (for these products, on average, 47% of embodied emissions are due to the electricity sector). So, to conclude, looking at tax rates at a product level, the main products affected would be the energy-intensive products, in particular when 16 The number in parenthesis after a product s name refers to the product s number in table 1. 12

13 imported from China, Indonesia, India, Russia, Korea, and Taiwan; the many emissions embodied in energy-intensive products coming from these countries are clearly related to the technology needed to produce them, but also, especially for China, India, and Russia, to the highly polluting electricity sector. Always at a product level, a slightly different picture is shown if, instead of considering only the taxation level obtained, we also consider the trade volume, by country and by product, between the regions analyzed and Europe. Figure 2 shows the 20 products, over the 308 analyzed, mostly affected by a system based on embodied emissions: these products would indeed bear more than the 60% of the total cost of the policy, represented by the size of each bubble, computed as the tax rates (shown in the horizontal axis) multiplied by the total value imported in Europe (shown in the vertical axis). 17 The main result that Figure 2 shows is that 14 out of 20 products would be imported from China that alone would sustain roughly 30% of the policy s cost. The ranking of these products seems to be more related with the volume of trade than the severity of the rates imposed. Figure products most affected by a system based on tax rates and trade volume ,0% 2,0% 4,0% 6,0% 8,0% 10,0% 12,0% 1.CHINA : Radio, television and comm. eq. 2.CHINA : Office machinery and computers 3.CHINA : Textiles 4. RUSSIA: Basic metals 5.UNITED STATES: Chemicals, chemical products 6.CHINA : Electrical machinery 7.CHINA : Machinery and equipment n.e.c. 8.CHINA : Chemicals, chemical products 9.CHINA : Fabricated metal products 10. RUSSIA: Coke, refined petroleum products 11.CHINA : Furniture; other manufactured goods 12.INDIA: Furniture; other manufactured goods 13.UNITED STATES: Other transport equipment 14.KOREA: Other transport equipment 15.CHINA : Rubber and plastic products 16.CHINA : Other non-metallic mineral products 17.CHINA : Medical and optical instruments 18.CHINA : Wearing apparel; furs 19.CHINA : Other transport equipment 20.CHINA : Basic metals The fifth and the thirteenth products that would bear most the cost of the would be chemical products (24) and other transport equipment (35) imported from USA: although tax rates would not be very high (respectively 1.9% and 1.1%), the volume of trade with Europe is very large. Conversely, very high tax rates more than the trade volume explain the cost the reform would implies on Russian products classified coke, refined petroleum products (23), and basic metals (27). Comparing the results obtained simulating s on embodied emissions with the results obtained by assuming s on avoided emissions, three main characteristics are worth being noticed. 17 The region that would actually bear the most part of a system is the region Rest of the World (ROW) that would pay roughly the 40% of the policy s cost. Anyway, we do not analyse the region ROW in detail because it aggregates several and different countries, and it would not be possible to provide a more detailed explanation to the results found. 13

14 First, the emissions Europe is avoiding, or the emissions that Europe would have produced if all products were made domestically, are on average very few. If we consider, for the 22 products analyzed, the emissions Europe would have generated producing all the inputs domestically, and we compare them with the emissions the other 14 regions are producing, out of 308 comparisons only in 12 cases the emissions embodied in foreign products would be less than the emissions avoided by Europe. This does not mean that Europe is actually polluting less than all the other countries or has the cleanest technology, 18 but it means that, if the EU produced domestically all the inputs imported, the environmental impact would be very low. Anyway, a second important result is that, in the context of the analysis on designs we propose, a comparison between rates obtained on embodied emissions and rates obtained considering avoided emissions has to be adjusted taking into account international differences in prices, since applying the same treatment to domestic products and to foreign products cannot depend on prices differences. The results shown in table 1 reveal that adjusting for differences in prices changes substantially the results: while the average rate on avoided emissions without deflating data would have been 0.9%, when data are deflated the average rate would be 1.4%. In any case, also taking into account international differences in prices, in general a system based on avoided emissions implies tariffs lower than in a system based on embodied emissions: only for 15% of the products analyzed (42 out of 308) a system based on avoided emissions would imply rates higher than a system based on embodied emissions. Finally, taking into account that 15% of goods that would be more affected by a system based on avoided emissions, the products that would be taxed more under a system based on avoided emissions are concentrated in few CPA categories, in particular tobacco products (16) imported from Brazil, Indonesia and Japan, textiles (17) from Brazil, Indonesia and Turkey, leather products (19) from Austria, Canada, and Turkey, and chemical products (24) imported from Brazil, Indonesia, and Turkey. This implies that these specific products imported from the listed countries are produced through very clean technologies able to produce a small amount of emissions compared with Europe. From these results it is also evident that s based on avoided emissions would be higher than s on embodied emissions manly in three countries: Brazil, Indonesia, and Turkey. In the following session we focus more in deep on specific countries to show the overall effect of the two tax designs for any of them Analysis at the country level Looking at the tax rates under the two regimes, the variance between the two approaches can be found at a country level too and also in this case with important differences across them, as shown in Figure 3, where countries are ordered based on the spreading of the on embodied emissions (Figure 3.a). For each country the label also shows, in parenthesis, the average tariff applied. For two countries, China and India, the differences between the two approaches would be very strong: considering embodied emissions, 100% of their products would be charged with tariffs higher than 2%, and the average tariff would be, respectively, 3.9% and 4.9%. Considering instead avoided 18 To know what country possesses the cleanest domestic technology we should compare European and foreign domestic emissions intensities, or European and foreign avoided emissions. In the comparison proposed here we instead consider European avoided emissions and foreign embodied emissions. 14

15 emissions only 27% of products would be mostly affected, and the average tariffs would be 1.9% and 1.7%. Figure 3. Percentage of products based on their tariff size, by country, under the two different s designs 100% 80% 60% 40% 20% 0% a. Embodied emissions 100% 80% 60% 40% 20% 0% b. Avoided emissions (adjusted for price differences) Source: own elaboration. tariff < 1% 1%< tariff <2% tariff >2% Although in a less decisive way, also for almost all the other countries analyzed a system based on embodied emissions would have a strong impact than s based on avoided emissions in both dimensions, the level of the rates and their spread across products, although the difference is less strong for Turkey, USA, Canada, Indonesia, Japan, and Australia. The country that performs differently from the rest of the regions analyzed is Brazil: indeed for this region a tariff system based on avoided emissions would be clearly worse than a system based on embodied emissions. In particular, 16 products (73%) would be taxed more under a system designed on avoided emissions. One of the reasons can be found in the contribution of the electricity sector to emissions: Brazil is indeed the country with the less polluting energy sector, producing on average, for each unit produced, only one fifth of the emissions provoked by the second less polluting country (Canada), and 80 times less than the most polluting one (India). Also analyzing the results at a county level it is interesting to show not only the effect in terms of rates, but also in terms of the cost that the two different designs would imply. Table 2 shows the cost of the two different designs for each country, considering in the first four columns the cost the policy would represent in relative terms, as a percentage of the value of manufactured products imported from any country by Europe, and also as a percentage of the total value of all products and services imported from each country by Europe. In the following columns Table 2 shows the share of the cost each country would bear. 15